SpeechConverterOffline
Defined in: SpeechConverterOffline.ts:12
SpeechConverter handles real-time speech-to-text conversion using the Whisper model. It manages audio input, preprocessing, and transcription directly in the browser.
Implements
Constructors
Constructor
new SpeechConverterOffline(): SpeechConverterOffline;
Defined in: SpeechConverterOffline.ts:23
Returns
SpeechConverterOffline
Methods
getStatus()
getStatus(): string;
Defined in: SpeechConverterOffline.ts:306
Retrieves the current status of the Whisper model.
Returns "loading" if the model has not been initialized yet,
otherwise returns the status string provided by the Whisper backend.
Returns
string
- The current operational status of the Whisper module.
Implementation of
SpeechConverterInterface.getStatus
getTextLog()
getTextLog(): string[];
Defined in: SpeechConverterOffline.ts:287
Returns
string[]
Implementation of
SpeechConverterInterface.getTextLog
getTranscribed()
getTranscribed(): string;
Defined in: SpeechConverterOffline.ts:241
Retrieves the latest transcription result from the Whisper model and logs it.
This method calls the underlying Whisper API to obtain the most recently transcribed text. If any text has been returned from whisper, it logs it.
Returns
string
- The transcribed text from the current audio chunk.
Throws
Throws if the Whisper module has not been initialized.
Implementation of
SpeechConverterInterface.getTranscribed
init()
init(modelPath, lang): Promise`<void>`;
Defined in: SpeechConverterOffline.ts:66
Initializes the Whisper module with the specified model and language.
This method:
- Creates the Whisper instance asynchronously.
- Loads the model file into the in-memory filesystem.
- Initializes Whisper with the model path and language code.
Parameters
| Parameter | Type | Description |
|---|---|---|
|
| Path or URL to the Whisper model file. |
|
| Language code (e.g., 'en') to configure the model. |
Returns
Promise<void>
- Resolves when the Whisper module is fully initialized.
Implementation of
startListening()
startListening(): void;
Defined in: SpeechConverterOffline.ts:152
Starts listening to the user's microphone input, collects audio chunks, and feeds them into the Whisper model for transcription in real time.
The method continuously gathers small chunks from AudioInputHandler,
combines them into fixed-size blocks, downsamples them to 16kHz (required by Whisper),
and sends them to the model for inference.
Returns
void
Throws
Throws if init() was not called before invoking this method.
Implementation of
SpeechConverterInterface.startListening
stopListening()
stopListening(): void;
Defined in: SpeechConverterOffline.ts:206
Stops the audio input stream and halts the real-time transcription process.
This should be called after startListening() to stop capturing microphone input
and free up system audio resources.
Returns
void
Throws
Throws if the Whisper module has not been initialized.